S01¶


Research	R0054 — Prompt Claims v2
Run	2026-03-31
Claim	C003
Search	S01

WebSearch — AI sycophancy and workflow compliance research

Summary¶

Field	Value
Source/Database	WebSearch
Query terms	AI sycophancy research workflow skipping compliance problem LLM; Anthropic Claude sycophancy paper AI agrees then ignores complex instructions research 2024 2025
Filters	None
Results returned	20 (two searches combined)
Results selected	4
Results rejected	16

Selected Results¶

Result	Title	URL	Rationale
S01-R01	Towards Understanding Sycophancy in Language Models (Anthropic)	https://www.anthropic.com/research/towards-understanding-sycophancy-in-language-models	Primary sycophancy research from Anthropic
S01-R02	Sycophancy in Large Language Models: Causes and Mitigations	https://arxiv.org/html/2411.15287v1	Comprehensive academic survey on sycophancy
S01-R03	When Helpfulness Backfires: LLMs and Misinformation Due to Sycophancy	https://pmc.ncbi.nlm.nih.gov/articles/PMC12045364/	Medical domain sycophancy with 100% compliance rates
S01-R04	The Yes-Machine Problem: Sycophantic AI Safety Crisis	https://www.webanditnews.com/2026/03/28/the-yes-machine-problem-how-sycophantic-ai-is-becoming-a-safety-crisis-nobody-wants-to-talk-about/	Recent overview of sycophancy as systemic problem

Rejected Results¶

Result	Title	URL	Rationale
S01-R05	Various sycophancy survey articles	Various	Duplicate coverage or insufficient depth
S01-R06	LLM evaluation papers	Various	About evaluation frameworks, not sycophancy behavior
S01-R07	Prompt injection articles	Various	About adversarial attacks, not sycophantic compliance
S01-R08	General AI safety articles	Various	Tangentially related but not specific to workflow compliance
S01-R09	Model system cards	Various	Documentation, not behavioral research
S01-R10	General LLM bias articles	Various	About content bias, not process compliance

Notes¶

Combined results from two related searches. The four selected sources provide comprehensive coverage of sycophancy from different angles: primary research (Anthropic), academic survey (arXiv), domain-specific evidence (medical), and current reporting.